Japanese Sentence Counter

Paste a paragraph of Japanese into a sentence counter built for English and you will usually get the answer 1 — because English tools split on the ASCII dot ., and Japanese ends its sentences with the small circle 。 (U+3002, maru) instead. Add the fact that Japanese writes no spaces between words, and most generic text statistics break down completely. This counter is built around Japanese punctuation conventions: it splits on 。, the full-width ！ and ？, tolerates mixed-script ASCII marks, and reports characters rather than unknowable “words”.

The terminators it recognises

Runs of terminators collapse into a single break: ！？, ？？？, and the doubled ellipsis …… (the conventional form in Japanese fiction — ellipses usually come in pairs) each end exactly one sentence rather than producing phantom empty ones. A segment must contain at least one kanji, kana, or alphanumeric character to count, so stray punctuation never inflates the total.

These conventions — which marks terminate, how they attach to the preceding character, and how they behave at line breaks — are codified in the W3C’s Japanese text layout requirements (JLReq), the standard reference for Japanese typography on the web.

Why the comma 、 must not split

The enumeration comma 、 (U+3001, tōten) separates clauses and list items inside a sentence, the way an English comma does:

春になると、桜が咲き、鳥が歌い始めます。
"When spring comes, the cherry blossoms bloom and the birds begin to sing."

Three clauses, two commas, one sentence — the only terminator is the final 。. A naive splitter that treats 、 as a boundary triples the count of typical prose. Japanese formal and legal writing habitually chains many 、-separated clauses before a single 。, so the error compounds in exactly the texts people most often need to measure.

Worked example

日本語は美しいです。話しますか？いいえ、少し難しいです……

1. 日本語は美しいです — ends at 。
2. 話しますか — ends at ？
3. いいえ、少し難しいです — ends at …… (two ellipsis characters, collapsed to one break; the 、 inside is ignored)

Sentence count: 3. With 27 counted characters of text, the average sentence length is 9 characters — short, conversational prose.

Characters, not words: the honest metric

Japanese is written wakachigaki-free — without spaces. Deciding where one word ends and the next begins requires a morphological analyser (MeCab, Sudachi, Kuromoji and similar tools exist precisely for this), and even those disagree on segmentation standards. Rather than pretend to a word count, this tool reports what is unambiguous:

• Character count with and without whitespace — the unit Japanese publishing actually uses. Manuscript lengths, translation quotes, and social-media limits in Japan are all specified in characters (字).
• Characters per sentence — the standard readability proxy. Journalistic Japanese tends to run noticeably shorter per sentence than legal or academic prose; if your average climbs past roughly 60-80 characters, readers will feel it.

Where sentence counts are actually used

• Manuscript and submission limits. Japanese essay contests, school assignments (原稿用紙 sheets hold 400 characters), and publishing briefs specify character budgets; sentence counts and per-sentence averages help you hit them without chopping meaning.
• Readability editing. Splitting one 120-character sentence into three 40-character sentences is the single most common edit in plain-Japanese (やさしい日本語) rewriting for public-sector communication.
• Translation and MT preparation. Sentence-segmenting a source text before machine translation improves quality; comparing source and target sentence counts verifies the segmentation survived.
• Subtitling and transcription QA. Expected-versus-counted sentence totals catch missing punctuation in transcripts quickly.

Edge cases and quirks

Quoted speech: a quotation ending in 」 typically has its terminator inside the brackets (「行きます。」) or omits it; the counter treats the terminator wherever it appears, and a bracket-closed segment without any terminator merges into the surrounding sentence — matching how editors count. Headings and list items usually carry no 。 at all (standard Japanese style), so they do not count as sentences unless they end in a terminator; add trailing marks if you want them counted. Half-width katakana and romaji mixed text is fine — ASCII . ! ? are accepted for the Latin islands inside Japanese text. The wave dash 〜 and prolonged sound mark ー never terminate a sentence and are treated as ordinary characters.

Everything runs locally in your browser — nothing you paste leaves your machine, so drafts, contracts, and unpublished manuscripts are safe.

Reference

• W3C — Requirements for Japanese Text Layout (JLReq)
• Unicode CJK Symbols and Punctuation block (U+3000–U+303F)